Evidence, Judgment and Accountability in Expected Credit Loss

A sophisticated Expected Credit Loss model can calculate probabilities to several decimal places. It can segment portfolios, estimate probability of default, loss given default and exposure at default, incorporate multiple economic scenarios and generate impairment figures with impressive mathematical precision. Yet mathematical sophistication alone does not establish that the resulting ECL is appropriate. The more important question is whether the institution can explain, substantiate and consistently defend why the reported number reasonably reflects the credit risk information available at the reporting date.

That question lies at the heart of credible implementation of IFRS 9. Expected Credit Loss is not intended to be a mechanical calculation detached from economic reality. It combines models, data, reasonable and supportable information, forward looking assessment, professional judgment and governance. The objective should therefore be neither to produce the lowest possible provision nor to produce the highest possible provision. The objective is to arrive at an unbiased and probability weighted estimate that appropriately reflects the requirements of the applicable financial reporting framework and the facts and circumstances existing at the relevant time.

A Model Is an Instrument, Not the Answer

At its simplest conceptual level, expected credit loss can be illustrated through three familiar components. These are Probability of Default, Loss Given Default and Exposure at Default. Their interaction helps explain how the likelihood of default, the severity of loss if default occurs and the amount exposed at the point of default collectively influence expected loss.

In practice, IFRS 9 measurement is considerably more sophisticated than a simple multiplication of these variables. Expected cash shortfalls must appropriately reflect probability weighted outcomes, reasonable and supportable information, relevant forward looking information and the time value of money. Lifetime ECL also requires consideration of credit risk over the relevant expected life of the financial instrument. Consequently, the apparent precision of a mathematical output should never be mistaken for certainty about the future.

Suppose two institutions hold economically similar portfolios of 100 billion monetary units. Their models generate initial ECL estimates of 4.0 billion and 4.1 billion respectively. The first institution subsequently applies a management adjustment of 1.0 billion because it identifies emerging sectoral deterioration that it considers inadequately captured by its existing model. The second institution concludes that no additional adjustment is necessary. Neither decision becomes inherently correct or incorrect simply because one institution reports a larger provision.

The appropriate questions concern the reasoning behind those decisions. What specific risk has been identified, why is that risk not adequately reflected in the existing model, what evidence supports the adjustment, how was its amount determined, whether possible duplication has been considered, what independent challenge occurred and under what circumstances the adjustment will eventually be modified or removed are all matters relevant to assessing the quality of the judgment. The credibility of an ECL outcome therefore depends not merely upon the final number but upon the integrity of the reasoning connecting risk, evidence, methodology and financial reporting.

The Evidence Chain

A sound ECL framework should establish a traceable relationship among economic conditions, observed credit risk signals, relevant data, assumptions, modelling decisions, adjustments, impairment outcomes, independent challenge, approval and disclosure. Each important connection should be capable of explanation and appropriate documentation.

If unemployment is expected to rise, management should be able to explain why that expectation is relevant to particular borrower segments and how the resulting risk is reflected in the ECL assessment. If property prices are expected to decline, the institution should consider the consequences for collateral dependent recoveries and loss severity. If a particular industry is deteriorating, management should determine whether the deterioration is already captured through probability of default, staging, loss given default, economic scenario weighting or another component before considering an additional adjustment.

This traceability matters because a financial model can otherwise create an illusion of certainty. A model may calculate impairment at 4.783 billion, but the presence of three decimal places does not transform an uncertain assumption into a certain fact. Numerical precision and economic reliability are related concepts, but they are not identical.

The Particular Importance of Stage 2

One of the most consequential areas of judgment under IFRS 9 concerns whether credit risk has increased significantly since initial recognition. The importance arises because the accounting consequence can be substantial. Broadly, Stage 1 exposures are subject to twelve month ECL, while Stage 2 exposures require lifetime ECL. Credit impaired financial assets fall within Stage 3 and are subject to the relevant requirements of IFRS 9.

The movement between stages can therefore materially alter impairment before an exposure becomes nonperforming. Significant Increase in Credit Risk should consequently be regarded as more than a technical classification exercise. Institutions require coherent criteria, relevant quantitative and qualitative indicators, appropriate consideration of applicable backstops and disciplined governance over exceptions and overrides.

A portfolio can therefore appear comparatively stable when examined only through the traditional NPL ratio while substantial deterioration is developing within Stage 2. This demonstrates why NPL and ECL should not be treated as interchangeable concepts. NPL generally identifies credit problems that have already manifested themselves according to the applicable regulatory definition, while ECL is forward looking and can recognize deterioration before conventional default or nonperformance occurs.

Management Adjustments Require Discipline

Management adjustments deserve particular attention because they occupy the important boundary between quantitative modelling and professional judgment. Such adjustments can be appropriate when relevant risks are not adequately captured by the core model. Structural economic changes, geopolitical disruption, unusual commodity movements, natural disasters, sectoral distress and limitations in available historical data can create circumstances in which models require supplementary judgment.

The existence of judgment, however, should not turn an adjustment into an unexplained balancing figure designed to produce a preferred financial result. A sound framework should identify the specific risk being addressed, explain the limitation of the existing model, establish an appropriate method of quantification, document relevant supporting information, consider possible duplication with risks already captured elsewhere, provide appropriate independent challenge and establish conditions under which the adjustment will be retained, revised, incorporated into the model or removed.

Neutrality is particularly important. An inadequately supported increase can distort financial performance just as an inadequately supported reduction can understate impairment. The purpose of ECL is not conservatism for its own sake or optimism for its own sake. Its purpose is faithful measurement within the applicable accounting requirements.

Where IFRS 9 Meets Basel III

Another important distinction concerns the relationship between accounting impairment and prudential regulation. IFRS 9 and the Basel capital framework interact, but they do not serve identical purposes. IFRS 9 addresses recognition and measurement for financial reporting, while the Basel framework and jurisdiction specific prudential requirements address capital adequacy and financial resilience.

An increase in accounting impairment can reduce profit and retained earnings and may therefore affect Common Equity Tier 1 capital, subject to applicable tax effects, regulatory adjustments, transitional arrangements and local prudential requirements. Capital consequences can subsequently influence distributions, balance sheet strategy, lending capacity and management decisions.

Consider a deliberately simplified illustration. Assume a bank has risk weighted assets of 100 billion monetary units and Common Equity Tier 1 capital of 13 billion. Its simplified Common Equity Tier 1 ratio is therefore 13 percent. If an additional ECL charge ultimately reduces Common Equity Tier 1 capital by 2 billion, while risk weighted assets remain unchanged and all other effects are ignored solely for illustration, Common Equity Tier 1 capital becomes 11 billion and the corresponding ratio becomes 11 percent.

The accounting estimate has consequently migrated through the financial statements into prudential capacity. This connection explains why effective ECL governance cannot reasonably remain confined to the finance function. Risk management, credit, finance, economics, data management, model validation, internal audit, senior management, relevant board committees and the board itself may each have distinct responsibilities within the governance architecture.

When NPL Reaches 30 Percent

The implications become particularly significant when an institution carries a high NPL ratio. Suppose an institution has total loans of 100 billion monetary units and an NPL ratio of 30 percent. Nonperforming loans therefore amount to 30 billion.

This does not mean that ECL automatically equals 30 billion, nor does it establish a universal provisioning percentage. Assume purely for illustration that the relevant defaulted portfolio has an economically supportable loss severity of 55 percent. A simplified calculation would produce an indicative loss amount of 16.5 billion. Actual IFRS 9 measurement may differ materially because expected recoverable cash flows, collateral values, realization costs, timing of recoveries, discounting, guarantees, portfolio characteristics and other relevant factors may affect the result.

There may also be material ECL within the remaining 70 billion of loans because performing does not mean risk free. Stage 1 exposures carry expected credit losses, while exposures experiencing significant increases in credit risk may require lifetime ECL in Stage 2. A high NPL ratio therefore reveals only part of the economic picture.

The distinction can be expressed metaphorically. The NPL ratio identifies rooms in a building where fire is already visible. ECL also asks whether smoke is accumulating elsewhere, whether electrical wiring is deteriorating and whether current conditions increase the probability of another fire. One measures visible damage while the other seeks, within the applicable framework, to measure both existing and reasonably expected credit deterioration.

The Capital Amplification Effect

The interaction among impairment, earnings and capital deserves particular attention. Suppose a bank earns 5 billion before impairment and recognizes 3 billion of impairment. Simplifying all other items, 2 billion remains before tax and other relevant adjustments. If deterioration subsequently causes impairment to increase to 7 billion, the same underlying operating performance produces a loss of 2 billion before tax and other adjustments.

That difference can affect retained earnings and regulatory capital. If capital falls while risk weighted assets remain relatively unchanged, regulatory capital ratios decline. If capital ratios approach applicable requirements or internal management buffers, the institution may need to retain earnings, raise capital, reduce risk weighted assets, improve recoveries, rebalance portfolios or adopt a combination of measures.

This illustrates the multiplier effect of deteriorating asset quality. Credit deterioration does not remain confined to the loan book. It can migrate from asset quality into impairment, from impairment into profitability, from profitability into capital and from capital into future lending capacity.

The Central Bank Dilemma and Procyclicality

Forward looking provisioning also raises an important macrofinancial issue. During an economic downturn, default expectations may increase while collateral values decline and economic forecasts weaken. Expected credit losses may therefore increase at precisely the time when profitability and capital are already under pressure.

If numerous institutions simultaneously respond by sharply restricting credit, the banking system can amplify an economic slowdown. Businesses experiencing weaker demand may find financing more difficult, investment may decline and additional borrowers may encounter repayment problems. This interaction is commonly discussed in the context of procyclicality.

The existence of this risk does not mean that credible impairment recognition should be suppressed. Delaying recognition may preserve reported profitability temporarily while allowing underlying vulnerabilities to accumulate. The broader policy challenge is therefore to maintain the integrity of financial reporting while addressing systemic resilience through appropriate prudential and macroprudential instruments within the mandate of the relevant authorities.

Accounting and prudential supervision should consequently be understood as complementary disciplines. Accounting seeks to represent financial reality according to applicable standards. Prudential regulation seeks to preserve institutional and systemic resilience according to the relevant legal and supervisory framework. Maintaining conceptual clarity between these purposes strengthens both.

The Boardroom Test

For directors and senior executives, perhaps the most useful question is not whether an ECL model is mathematically perfect. No complex model can eliminate uncertainty from the future. A more useful governance question is whether the institution could reconstruct, at a later date, the evidence, assumptions, judgments and approval process that existed when the ECL estimate was determined.

This perspective changes the boardroom conversation. Attention moves from defending a number toward demonstrating the integrity of the process through which the number was produced. Directors do not need to become statistical modellers, but effective oversight requires sufficient understanding to challenge material assumptions, recognize significant model limitations, assess whether management judgments have received appropriate independent scrutiny and determine whether conclusions are supported by relevant information.

The objective should not be to create a model that no regulator, auditor or independent reviewer can question. Questioning is an essential part of sound governance. The objective is to establish a framework capable of responding to legitimate questions through evidence, transparent reasoning and appropriate documentation.

From Model Governance to Institutional Credibility

A mature ECL framework should operate as a continuing process of measurement, challenge, validation, monitoring, back testing, recalibration, documentation and governance. These activities allow institutions to determine whether assumptions remain relevant as economic conditions and portfolio characteristics change.

Back testing is especially important. If realized defaults, recoveries or losses repeatedly differ materially from previous estimates, management should investigate the reasons and consider whether methodologies or assumptions require refinement. Models should learn from evidence rather than merely survive periodic validation.

The same reasoning applies to management adjustments. An adjustment originally introduced to address a temporary model limitation should not automatically become permanent. If the same adjustment remains material over several reporting periods, management should consider whether the underlying risk should be incorporated more directly into the core methodology, subject to applicable requirements and appropriate validation.

The Measure of a Strong ECL Framework

The strongest ECL framework is not necessarily the most complicated framework, nor is it necessarily the one producing the highest or lowest provision. Its strength lies in whether relevant data are sufficiently reliable, assumptions are reasonable and supportable, methodologies are appropriate, significant judgments are transparent, limitations are understood, adjustments are controlled, outcomes are tested and responsibilities are clearly governed.

There is therefore an important distinction between model sophistication and institutional maturity. A highly sophisticated model can operate within weak governance, while an appropriately designed model supported by strong data, disciplined validation, credible challenge and effective oversight can provide a sound basis for financial reporting and risk management.

IFRS 9 ultimately asks financial institutions to confront an unavoidable reality. The future cannot be known with certainty, but uncertainty can be approached with discipline. ECL should therefore not become a contest between management, auditors and regulators, nor should it become an exercise in producing a predetermined accounting result. Its enduring value lies in creating a common analytical language through which emerging credit risk can be identified, measured, challenged and communicated responsibly.

A credible ECL framework ultimately rests upon four connected foundations. The model provides measurement, evidence provides credibility, governance provides accountability and transparency provides trust. When these foundations reinforce one another, ECL becomes more than an accounting calculation. It becomes part of the institutional architecture through which a bank understands risk before that risk becomes loss.

Author’s Note: This article presents an independent and general discussion of financial reporting, credit risk management and prudential concepts. It does not assess or comment upon any particular financial institution, regulator, jurisdiction, supervisory action or individual ECL methodology. Actual accounting treatment, impairment measurement, capital treatment and supervisory requirements depend upon the applicable standards, laws, regulations and the specific facts and circumstances. Institution specific conclusions should therefore be determined by reference to the applicable requirements and appropriate professional assessment.